Electronic Journal of Qualitative Theory of Differential Equations 2012, No. 38, 1-34;http://www.math.u-szeged.hu/ejqtde/
Two-Point Boundary Value Problems For Strongly Singular Higher-Order Linear Differential Equations
With Deviating Arguments
Sulkhan Mukhigulashvili
∗and Nino Partsvania
Abstract
For strongly singular higher-order differential equations with deviating argu- ments, under two-point conjugated and right-focal boundary conditions, Agarwal- Kiguradze type theorems are established, which guarantee the presence of Fred- holm’s property for the above mentioned problems. Also we provide easily verifi- able best possible conditions that guarantee the existence of a unique solution of the studied problems.
2000 Mathematics Subject Classification: 34K06, 34K10
Key words and phrases: Higher order differential equation, linear, deviating argu- ment, strong singularity, Fredholm’s property.
1 Statement of the main results
1.1. Statement of the problems and the basic notations. Consider the differential equations with deviating arguments
u(n)(t) = Xm
j=1
pj(t)u(j−1)(τj(t)) +q(t) for a < t < b, (1.1) with the two-point boundary conditions
u(i−1)(a) = 0 (i= 1,· · ·, m), u(j−1)(b) = 0 (j = 1,· · · , n−m), (1.2) u(i−1)(a) = 0 (i= 1,· · · , m), u(j−1)(b) = 0 (j =m+ 1,· · · , n). (1.3) Here n≥2, m is the integer part of n/2, −∞< a < b < +∞, pj, q ∈Lloc(]a, b[) (j = 1,· · · , m), and τj :]a, b[→]a, b[ are measurable functions. By u(j−1)(a) (u(j−1)(b)) we denote the right (the left) limit of the function u(j−1) at the point a(b). Problems (1.1), (1.2), and (1.1), (1.3) are said to be singular if some or all the coefficients of (1.1) are non-integrable on [a, b], having singularities at the end-points of this segment.
∗Corresponding author.
The linear ordinary differential equations and differential equations with deviating arguments with boundary conditions (1.2) and (1.3), and with the conditions
Zb a
(s−a)n−1(b−s)2m−1[(−1)n−mp1(s)]+ds <+∞, Zb
a
(s−a)n−j(b−s)2m−j|pj(s)|ds <+∞ (j = 2,· · · , m), Zb
a
(s−a)n−m−1/2(b−s)m−1/2|q(s)|ds <+∞,
(1.4)
and Zb
a
(s−a)n−1[(−1)n−mp1(s)]+ds <+∞, Zb
a
(s−a)n−j|pj(s)|ds <+∞ (j = 2,· · · , m), Zb
a
(s−a)n−m−1/2|q(s)|ds <+∞,
(1.5)
respectively, were studied by I. Kiguradze, R. P. Agarwal and some other authors (see [1], [2], [4] - [22]).
The first step in studying the linear ordinary differential equations under conditions (1.2) or (1.3), in the case when the functions pj and q have strong singularities at the points a and b, i.e. when conditions (1.4) and (1.5) are not fulfilled, was made by R. P.
Agarwal and I. Kiguradze in the article [3].
In this paper the Agarwal-Kiguradze type theorems are proved which guarantee Fred- holm’s property for problems (1.1), (1.2), and (1.1), (1.3) (see Definition 1.1). Moreover, we establish optimal, in some sense, sufficient conditions for the solvability of problems (1.1), (1.2), and (1.1), (1.3).
Throughout the paper we use the following notation.
R+= [0,+∞[;
[x]+ is the positive part of numberx, that is [x]+ = x+|x|2 ;
Lloc(]a, b[) (Lloc(]a, b])) is the space of functions y:]a, b[→ R, which are integrable on [a+ε, b−ε]; ([a+ε, b]) for arbitrary small ε >0;
Lα,β(]a, b[) (L2α,β(]a, b[)) is the space of integrable (square integrable) with the weight (t−a)α(b−t)β functionsy :]a, b[→R, with the norm
||y||Lα,β = Zb
a
(s−a)α(b−s)β|y(s)|ds
||y||L2α,β =Zb
a
(s−a)α(b−s)βy2(s)ds1/2
; L([a, b]) =L0,0(]a, b[), L2([a, b]) =L20,0(]a, b[);
M(]a, b[) is the set of measurable functions τ :]a, b[→]a, b[;
Le2α,β(]a, b[) (Le2α(]a, b]) is the Banach space of functions y ∈ Lloc(]a, b[) (Lloc(]a, b])), satisfying
µ1 ≡maxnhZt
a
(s−a)αZt
s
y(ξ)dξ2
dsi1/2
:a≤t≤ a+b 2
o +
+ maxnhZb
t
(b−s)βZs
t
y(ξ)dξ2
dsi1/2
: a+b
2 ≤t≤bo
<+∞,
µ2 ≡maxnhZt
a
(s−a)αZt
s
y(ξ)dξ2
dsi1/2
:a≤t≤bo
<+∞. The norm in this space is defined by the equality || · ||Le2α,β =µ1 (|| · ||Le2α =µ2).
Cen−1, m(]a, b[) (Cen−1, m(]a, b])) is the space of functions y ∈ Celocn−1(]a, b[) (y∈Celocn−1(]a, b])), satisfying
Zb a
|y(m)(s)|2ds <+∞. (1.6)
When problem (1.1), (1.2) is discussed, we assume that for n= 2m, the conditions pj ∈Lloc(]a, b[) (j = 1,· · · , m) (1.7) are fulfilled, and for n = 2m+ 1,along with (1.7), the conditions
lim sup
t→b
(b−t)2m−1 Zt t1
p1(s)ds
<+∞ (t1 = a+b
2 ) (1.8)
are fulfilled. Problem (1.1), (1.3) is discussed under the assumptions
pj ∈Lloc(]a, b]) (j = 1,· · ·, m). (1.9) A solution of problem (1.1), (1.2) ((1.1), (1.3)) is sought in the space Cen−1, m(]a, b[) (Cen−1, m(]a, b])).
By hj :]a, b[×]a, b[→R+ and fj :R×M(]a, b[)→Cloc(]a, b[×]a, b[) (j = 1, . . . , m) we denote the functions and the operators, respectively, defined by the equalities
h1(t, s) =
Zt s
(ξ−a)n−2m[(−1)n−mp1(ξ)]+dξ ,
hj(t, s) =
Zt s
(ξ−a)n−2mpj(ξ)dξ
(j = 2,· · · , m),
(1.10)
and,
fj(c, τj)(t, s) =
Zt s
(ξ−a)n−2m|pj(ξ)|
τZj(ξ) ξ
(ξ1−c)2(m−j)dξ1
1/2dξ
(j = 1,· · · , m). (1.11)
Let, moreover,
m!! =
(1 for m≤0 1·3·5· · ·m for m≥1, if m= 2k+ 1.
1.2. Fredholm type theorems.
Along with (1.1), we consider the homogeneous equation v(n)(t) =
Xm j=1
pj(t)v(j−1)(τj(t)) for a < t < b. (1.10) In the case where conditions (1.4) and (1.5) are violated, the question on the presence of the Fredholm’s property for problem (1.1), (1.2) ((1.1), (1.3)) in some subspace of the space Celocn−1,m(]a, b[) (Celocn−1,m(]a, b])) remains so far open. This question is answered in Theorem 1.1 (Theorem 1.2 ) formulated below which contains optimal in a certain sense conditions guaranteeing the Fredholm’s property for problem (1.1), (1.2) ((1.1), (1.3)) in the space Cen−1, m(]a, b[) (Cen−1, m(]a, b])).
Definition 1.1. We will say that problem (1.1), (1.2) ((1.1), (1.3)) has the Fredholm’s property in the space Cen−1,m(]a, b[) (Cen−1,m(]a, b])), if the unique solvability of the cor- responding homogeneous problem (1.10), (1.2) ((1.10), (1.3)) in that space implies the unique solvability of problem (1.1), (1.2) ((1.1), (1.3)) for every q∈Le22n−2m−2,2m−2(]a, b[) (q∈Le22n−2m−2(]a, b])).
Theorem 1.1. Let there exist a0 ∈]a, b[, b0 ∈]a0, b[, numbers lkj > 0, γkj > 0, and functions τj ∈M(]a, b[) (k = 0,1, j = 1, . . . , m) such that
(t−a)2m−jhj(t, s)≤l0j for a < t≤s≤a0, lim sup
t→a
(t−a)m−12−γ0jfj(a, τj)(t, s)<+∞, (1.12) (b−t)2m−jhj(t, s)≤l1j for b0 ≤s ≤t < b,
lim sup
t→b
(b−t)m−12−γ1jfj(b, τj)(t, s)<+∞, (1.13)
and Xm
j=1
(2m−j)22m−j+1
(2m−1)!!(2m−2j+ 1)!! lkj <1 (k= 0,1). (1.14)
Let, moreover,(1.10), (1.2)have only the trivial solution in the spaceCen−1,m(]a, b[). Then problem (1.1), (1.2) has the unique solution u for every q ∈ Le22n−2m−2,2m−2(]a, b[), and there exists a constant r, independent of q, such that
||u(m)||L2 ≤r||q||Le22n−2m−2,2m−2. (1.15) Corollary 1.1. Let numbers κkj, νkj ∈R+ be such that
νk1 >2n+ 2−2k(2m−n), νkj >2 (k = 0,1; j = 2, . . . , m), (1.16) lim sup
t→a
|τj(t)−t|
(t−a)ν0j <+∞, lim sup
t→b
|τj(t)−t|
(b−t)ν1j <+∞, (1.17)
and Xm
j=1
22m−j+1
(2m−1)!!(2m−2j+ 1)!!κkj <1 (k = 0,1). (1.18) Moreover, let κ∈R+, p0j ∈Ln−j,2m−j(]a, b[;R+), and
− κ
[(t−a)(b−t)]2n −p01(t)≤(−1)n−mp1(t)≤
≤ κ01
(t−a)n + κ11
(t−a)n−2m(b−t)2m +p01(t), (1.19)
|pj(t)| ≤ κ0j
(t−a)n−j+1 + κ1j
(t−a)n−2m(b−t)2m−j+1 +p0j(t) (j = 2, . . . , m). (1.20) Let, moreover,(1.10), (1.2)have only the trivial solution in the spaceCen−1,m(]a, b[). Then problem (1.1), (1.2) has the unique solution u for every q ∈ Le22n−2m−2,2m−2(]a, b[), and there exists a constant r, independent of q, such that (1.15) holds.
Theorem 1.2. Let there exist a0 ∈]a, b[, numbers l0j > 0, γ0j > 0, and functions τj ∈ M(]a, b[) such that condition (1.12) is fulfilled and
Xm j=1
(2m−j)22m−j+1
(2m−1)!!(2m−2j+ 1)!!l0j <1. (1.21) Let, moreover, problem (1.10), (1.3)have only the trivial solution in the spaceCen−1,m(]a, b]).
Then problem (1.1), (1.3) has the unique solution u for every q ∈ Le22n−2m−2(]a, b]), and there exists a constant r, independent of q, such that
||u(m)||L2 ≤r||q||Le22n−2m−2. (1.22) Corollary 1.2. Let numbers κ0j, ν0j ∈R+ be such that
ν01>2n+ 2, ν0j ≥2 (j = 2, . . . , m), (1.23) lim sup
t→a
|τj(t)−t|
(t−a)ν0j <+∞, (1.24)
and m X
j=1
22m−j+1
(2m−1)!!(2m−2j+ 1)!!κ0j <1. (1.25) Let, moreover, κ∈R+, p0j ∈Ln−j,0(]a, b];R+), and
− κ
(t−a)2n −p01(t)≤(−1)n−mp1(t)≤ κ01
(t−a)n +p01(t), (1.26)
|pj(t)| ≤ κ0j
(t−a)n−j+1 +p0j(t) (j = 2, . . . , m). (1.27) Let, moreover, problem (1.10), (1.3)have only the trivial solution in the spaceCen−1,m(]a, b]).
Then problem (1.1), (1.3) has the unique solution u for every q ∈ Le22n−2m−2(]a, b]), and there exists a constant r, independent of q, such that (1.22) holds.
Theorem 1.3. Let c1 =a, c2 =b, ess sup
a<t<b
1
|t−ci|m+1−j
τj(t)
Z
t
|ξ−ci|m−j−1dξ
<+∞ (j = 1, . . . , m) (1.28)
if i= 1,2 (if i= 1),
pj ∈Ln−j,2m−j(]a, b[)
pj ∈Ln−j,0(]a, b])
(j = 1, . . . , m), (1.29) and let problem (1.1), (1.2) ((1.1), (1.3))be uniquely solvable in the space Cen−1, m(]a, b[) (in the spaceCen−1, m(]a, b]). Then this problem is uniquely solvable in the spaceCen−1(]a, b[) (in the space Cen−1(]a, b]) as well.
Remark 1.1. In [3], an example is constructed which demonstrates that if condition (1.29) is violated, then problem (1.1), (1.2) (problem (1.1), (1.3)) with τj(t) ≡ t(j = 1, . . . , m) may be uniquely solvable in the spaceCen−1,m(]a, b[) (in the spaceCen−1,m(]a, b])) and this problem may have infinite set of solutions in the space Celoc(]a, b[) (in the space Celoc(]a, b])).
Also, in [3] it is demonstrated that strict inequalities (1.14), (1.21), (1.18), (1.25) are sharp because they cannot be replaced by nonstrict ones.
1.2. Existence and uniqueness theorems.
Theorem 1.4. Let there exist numbers t∗ ∈]a, b[, ℓkj > 0, lkj ≥ 0, and γkj > 0 (k = 0,1;j = 1, . . . , m) such that along with
Xm j=1
(2m−j)22m−j+1l0j
(2m−1)!!(2m−2j+ 1)!! + 22m−j−1(t∗−a)γ0jl0j
(2m−2j−1)!!(2m−3)!!p 2γ0j
< 1
2, (1.30) Xm
j=1
(2m−j)22m−j+1l1j
(2m−1)!!(2m−2j+ 1)!! + 22m−j−1(b−t∗)γ0jl1j
(2m−2j−1)!!(2m−3)!!p 2γ1j
< 1
2, (1.31)
the conditions
(t−a)2m−jhj(t, s)≤l0j, (t−a)m−γ0j−1/2fj(a, τj)(t, s)≤l0j for a < t≤s≤t∗, (1.32) (b−t)2m−jhj(t, s)≤l1j, (b−t)m−γ1j−1/2fj(b, τj)(t, s)≤l1j for t∗≤s≤t < b (1.33) hold. Then for everyq ∈Le22n−2m−2,2m−2(]a, b[)problem (1.1), (1.2) is uniquely solvable in the space Cen−1, m(]a, b[).
To illustrate this theorem, we consider the second order differential equation with a deviating argument
u′′(t) =p(t)u(τ(t)) +q(t), (1.34) under the boundary conditions
u(a) = 0, u(b) = 0. (1.35)
From Theorem 1.4, with n = 2, m = 1, t∗ = (a+b)/2, γ01 =γ11 = 1/2, l01 = l11 = κ0, l01=l11 =√
2κ1/√
b−a, we get
Corollary 1.3. Let function τ ∈M(]a, b[) be such that 0≤τ(t)−t≤ 26
(b−a)6(t−a)7 for a < t≤ a+b 2 ,
− 26
(b−a)6(b−t)7 ≤t−τ(t)≤0 for a+b
2 ≤t < b.
(1.36)
Moreover, let function p:]a, b[→R and constants κ0, κ1 be such that
− 2−2(b−a)2κ0
[(b−t)(t−a)]2 ≤p1(t)≤ 2−7(b−a)6κ1
[(b−t)(t−a)]4 for a < t≤b (1.37) and
4κ0 +κ1 < 1
2. (1.38)
Then for every q ∈ Le20,0(]a, b[) problem (1.34), (1.35) is uniquely solvable in the space Ce1,1(]a, b[).
Theorem 1.5. Let there exist numbers t∗ ∈]a, b[, ℓ0j > 0, ℓ0j ≥ 0, and γ0j > 0 (j = 1, . . . , m) such that conditions
(t−a)2m−jhj(t, s)≤l0j, (t−a)m−γ0j−1/2fj(a, τj)(t, s)≤l0j for a < t≤s≤b, (1.39) and
Xm j=1
(2m−j)22m−j+1l0j
(2m−1)!!(2m−2j+ 1)!! + 22m−j−1(t∗−a)γ0jl0j (2m−2j−1)!!(2m−3)!!p
2γ0j
<1 (1.40) hold. Then for every q ∈Le22n−2m−2(]a, b]) problem (1.1), (1.3) is uniquely solvable in the space Cen−1, m(]a, b]).
Theorem 1.6. Let there exist numbers t∗ ∈]a, b[, ℓkj > 0, lkj ≥ 0, and γkj > 0 (k = 0,1;j = 1, . . . , m) such that along with (1.40) and
Xm j=1
(2m−j)22m−j+1l1j
(2m−1)!!(2m−2j + 1)!!+ 22m−j−1(b−t∗)γ0jl1j (2m−2j−1)!!(2m−3)!!p
2γ1j
<1, (1.41) conditions (1.32), (1.33) hold. Moreover, let τj ∈M(]a, b[) (j = 1, . . . , n) and
sign[(τj(t)−t∗)(t−t∗)]≥0 for a < t < b. (1.42) Then for every q ∈ Le22n−2m−2,2m−2(]a, b[) problem (1.1), (1.2) is uniquely solvable in the space Cen−1, m(]a, b[).
Also, from Theorem 1.6, with n = 2, m = 1, t∗ = (a+b)/2, γ01 = γ11 = 1/2, l01 = l11 =κ0, l01=l11=√
2κ1/√
b−a, we get
Corollary 1.4. Let functions p:]a, b[→R, τ ∈M(]a, b[) and constants κ0 >0, κ1 >0 be such that along with (1.36) and (1.37) the inequalities
sign[(τ(t)− a+b
2 )(t−a+b
2 )]≥0 for a < t < b (1.43) and
4κ0+κ1 <1 (1.44)
hold. Then for every q ∈ Le20,0(]a, b[) problem (1.34), (1.35) is uniquely solvable in the space Ce1,1(]a, b[).
2 Auxiliary propositions
2.1. Lemmas on integral inequalities. Now we formulate two lemmas which are proved in [3].
Lemma 2.1. Let ∈Celocm−1(]t0, t1[) and
u(j−1)(t0) = 0 (j = 1, . . . , m),
t1
Z
t0
|u(m)(s)|2ds <+∞. (2.1)
Then Zt t0
(u(j−1)(s))2
(s−t0)2m−2j+2ds≤ 2m−j+1 (2m−2j+ 1)!!
2Zt t0
|u(m)(s)|2ds for t0 ≤t≤t1. (2.2)
Lemma 2.2. Let u∈Celocm−1(]t0, t1[), and u(j−1)(t1) = 0 (j = 1, . . . , m),
t1
Z
t0
|u(m)(s)|2ds <+∞. (2.3) Then
t1
Z
t
(u(j−1)(s))2
(t1−s)2m−2j+2ds≤ 2m−j+1 (2m−2j+ 1)!!
2Zt1
t
|u(m)(s)|2ds for t0 ≤t≤t1. (2.4) Let t0, t1 ∈]a, b[, u∈Celocm−1(]t0, t1[) and τj ∈M(]a, b[) (j = 1, . . . , m).Then we define the functions µj : [a, (a+b)/2]×[(a+b)/2, b]×[a, b]→ [a, b], ρk : [t0, t1]→ R+(k = 0,1), λj : [a, b]×]a, (a+b)/2]×[(a+b)/2, b[×]a, b[→R+, by the equalities
µj(t0, t1, t) =
τj(t) for τj(t)∈[t0, t1] t0 for τj(t)< t0
t1 for τj(t)> t1
,
ρk(t) =
tk
Z
t
|u(m)(s)|2ds
, λj(c, t0, t1, t) =
µj(tZ0,t1,t) t
(s−c)2(m−j)ds
1/2.
(2.5)
Let also functions αj : R+3 ×[0,1[→ R+ and βj ∈ R+×[0,1[→ R+ (j = 1,· · · , m) be defined by the equalities
αj(x, y, z, γ) =x+ 2m−jy zγ
(2m−2j−1)!!, βj(y, γ) = 22m−j−1
(2m−2j−1)!!(2m−3)!!
yγ
√2γ. (2.6) Lemma 2.3. Let a0 ∈]a, b[, t0 ∈]a, a0[, t1 ∈]a0, b[, and the function u ∈ Celocm−1(]t0, t1[) be such that conditions (2.1) hold. Moreover, let constants l0j >0, l0j ≥0, γ0j >0, and functions pj ∈Lloc(]t0, t1[), τj ∈M(]a, b[) be such that the inequalities
(t−t0)2m−1
a0
Z
t
[p1(s)]+ds≤l0 1, (2.7)
(t−t0)2m−j
a0
Z
t
pj(s)ds
≤l0j (j = 2, . . . , m), (2.8)
(t−t0)m−12−γ0j
a0
Z
t
pj(s)λj(t0, t0, t1, s)ds
≤l0j (j = 1, . . . , m) (2.9) hold for t0 < t≤a0. Then
a0
Z
t
pj(s)u(s)u(j−1)(µj(t0, t1, s))ds≤
≤αj(l0j, l0j, a0−a, γ0j)ρ1/20 (τ∗)ρ1/20 (t) +l0jβj(a0−a, γ0j)ρ1/20 (τ∗)ρ1/20 (a0)+
+l0j
(2m−j)22m−j+1
(2m−1)!!(2m−2j+ 1)!!ρ0(a0) for t0 < t≤a0,
(2.10)
where τ∗ = sup{µj(t0, t1, t) :t0 ≤t ≤a0, j = 1, . . . , m} ≤t1.
Proof. In view of the formula of integration by parts, fort ∈[t0, a0] we have
a0
Z
t
pj(s)u(s)u(j−1)(µj(t0, t1, s))ds=
a0
Z
t
pj(s)u(s)u(j−1)(s)ds+
+
a0
Z
t
pj(s)u(s) µj(tZ0,t1,s)
s
u(j)(ξ)dξ
ds =u(t)u(j−1)(t)
a0
Z
t
pj(s)ds+
+ X1 k=0
a0
Z
t
Za0
s
pj(ξ)dξ
u(k)(s)u(j−k)(s)ds+
a0
Z
t
pj(s)u(s) µj(tZ0,t1,s)
s
u(j)(ξ)dξ ds
(2.11)
(j = 2, . . . , m), and
a0
Z
t
p1(s)u(s)u(µ1(t0, t1, s))ds≤
a0
Z
t
[p1(s)]+u2(s)ds+
+
a0
Z
t
|p1(s)u(s)|
µ1(tZ0,t1,s) s
u′(ξ)dξ
ds≤u2(t)
a0
Z
t
[p1(s)]+ds+
+2
a0
Z
t
Za0
s
[p1(ξ)]+dξ
|u(s)u′(s)|ds+
a0
Z
t
|p1(s)u(s)|
µ1(tZ0,t1,s) s
u′(ξ)dξ ds.
(2.12)
On the other hand, by conditions (2.1), the Schwartz inequality and Lemma 2.1, we deduce that
|u(j−1)(t)|= 1 (m−j)!
Zt t0
(t−s)m−ju(m)(s)ds
≤(t−t0)m−j+1/2ρ1/20 (t) (2.13) for t0 ≤t≤a0 (j = 1, . . . , m).If along with this, in the case j > 1, we take into account inequality (2.8), and lemma 2.1, for t ∈[t0, a0], we obtain the estimates
u(t)u(j−1)(t)
a0
Z
t
pj(s)ds
≤(t−t0)2m−j
a0
Z
t
pj(s)ds
ρ0(t)≤l0jρ0(t), (2.14)
and X1 k=0
a0
Z
t
Za0
s
pj(ξ)dξ
u(k)(s)u(j−k)(s)ds ≤l0j
X1 k=0
a0
Z
t
|u(k)(s)u(j−k)(s)| (s−t0)2m−j ds≤
≤ l0j
X1 k=0
Za0
t
|u(k)(s)|2ds (s−t0)2m−2k
1/2Za0
t
|u(j−k)(s)|2ds (s−t0)2m+2k−2j
1/2
≤
≤l0jρ0(a0) X1
k=0
22m−j
(2m−2k−1)!!(2m+ 2k−2j−1)!!.
(2.15)
Analogously, if j = 1, by (2.7) we obtain u2(t)
a0
Z
t
[p1(s)]+ds ≤l01ρ0(t),
2
a0
Z
t
Za0
s
[p1(ξ)]+dξ
|u(s)u′(s)|ds≤l01ρ0(a0)(2m−1)22m [(2m−1)!!]2
(2.16)
for t0 < t≤a0.
By the Schwartz inequality, Lemma 2.1, and the fact thatρ0 is nondecreasing function, we get
µj(tZ0,t1,s) s
u(j)(ξ)dξ
≤ 2m−j
(2m−2j−1)!!λj(t0, t0, t1, s)ρ1/20 (τ∗) (2.17) for t0 < s≤a0. Also, due to (2.2), (2.9) and (2.13), we have
|u(t)|
a0
Z
t
|pj(s)|λj(t0, t0, t1, s)ds= (t−t0)m−1/2ρ1/20 (t)
a0
Z
t
|pj(s)|λj(t0, t0, t1, s)ds≤
≤l0j(t−t0)γ0jρ1/20 (t),
a0
Z
t
|u′(s)|Za0
s
|pj(ξ)|λj(t0, t0, t1, ξ)dξ
ds ≤l0j a0
Z
t
|u′(s)|
(s−t0)m−12−γ0jds≤
≤l0j
2m−1(a0−a)γ0j (2m−3)!!p
2γ0j
ρ1/20 (a0) for t0 < t≤a0.From the last three inequalities it is clear that
(2m−2j −1)!!
2m−jρ1/20 (τ∗)
a0
Z
t
pj(s)u(s) µj(tZ0,t1,s)
s
u(j)(ξ)dξ ds
≤
a0
Z
t
|pj(s)u(s)|λj(t0, t0, t1, s)ds≤
≤ |u(t)|
a0
Z
t
|pj(s)|λj(t0, t0, t1, s)ds+
a0
Z
t
|u′(s)|Za0
s
|pj(ξ)|λj(t0, t0, t1, ξ)dξ ds≤
≤l0j(t−t0)γ0jρ1/20 (t) +l0j
2m−1(a0−a)γ0j (2m−3)!!p
2γ0j
ρ1/20 (a0)
(2.18)
for t0 < t ≤ a0. Now, note that from (2.11) and (2.12) by (2.14)-(2.16) and (2.18), it immediately follows inequality (2.10).
The following lemma can be proved similarly to Lemma 2.3.
Lemma 2.4. Let b0 ∈]a, b[, t1 ∈]b0, b[, t0 ∈]a, b0[, and the function u∈Celocm−1(]t0, t1[) be such that conditions (2.3) hold. Moreover, let constants l1j > 0, l1j ≥ 0, γ1j > 0, and functions pj ∈Lloc(]t0, t1[), τj ∈M(]a, b[) be such that the inequalities
(t1−t)2m−1 Zt b0
[p1(s)]+ds≤l1 1, (2.19)
(t1−t)2m−j
Zt b0
pj(s)ds
≤l1j (j = 2, . . . , m), (2.20)
(t1−t)m−12−γ1j
Zt b0
pj(s)λj(t1, t0, t1, s)ds
≤l1j (j = 1, . . . , m) (2.21)
hold for b0 < t≤t1. Then Zt b0
pj(s)u(s)u(j−1)(µj(t0, t1, s))ds≤
≤αj(l1j, l1j, b−b0, γ1j)ρ1/21 (τ∗)ρ1/21 (t) +l1jβj(b−b0, γ1j)ρ1/21 (τ∗)ρ1/21 (b0)+
+l1j
(2m−j)22m−j+1
(2m−1)!!(2m−2j+ 1)!!ρ1(b0) for b0 ≤t < t1,
(2.22)
where τ∗ = inf{µj(t0, t1, t) :b0 ≤t≤t1, j = 1, . . . , m} ≥t0.
2.2. Lemma on the property of functions from the space Cen−1,m(]a, b[).
Lemma 2.5. Let
w(t) =
n−mX
i=1 n−mX
k=i
cik(t)u(n−k)(t)u(i−1)(t),
where Cen−1,m(]a, b[), and each cik : [a, b] → R is an (n−k −i+ 1) -times continuously differentiable function. Moreover, if
u(i−1)(a) = 0 (i= 1, . . . , m), lim sup
t→a
|cii(t)|
(t−a)n−2m <+∞ (i= 1, . . . , n−m), then
lim inf
t→a |w(t)|= 0, and if u(i−1)(b) = 0 (i= 1, . . . , n−m), then
lim inf
t→b |w(t)|= 0.
The proof of this lemma is given in [9].
2.3. Lemmas on the sequences of solutions of auxiliary problems.
Now for every natural k we consider the auxiliary boundary problems u(n)(t) =
Xm j=1
pj(t)u(j−1)(µj(t0k, t1k, t)) +qk(t) for t0k ≤t≤t1k, (2.23) u(i−1)(t0k) = 0 (i= 1, . . . , m), u(j−1)(t1k) = 0 (j = 1, . . . , n−m), (2.24) where
a < t0k < t1k < b (k∈N), lim
k→+∞t0k =a, lim
k→+∞t1k=b, (2.25) and
u(n)(t) = Xm
j=1
pj(t)u(j−1)(µj(t0k, b, t)) +qk(t) for t0k ≤t≤b, (2.26) u(i−1)(t0k) = 0 (i= 1, . . . , m), u(j−1)(b) = 0 (j = 1, . . . , n−m), (2.27) where
a < t0k< b (k ∈N), lim
k→+∞t0k =a. (2.28)
Throughout this section, when problems (1.1), (1.2) and (2.23), (2.24) are discussed we assume that
pj ∈Lloc(]a, b[) (j = 1, ..., m), q, qk∈Le22n−2m−2,2m−2(]a, b[), (2.29) and for an arbitrary (m−1)-times continuously differentiable function x :]a, b[→ R, we set
Λk(x)(t) = Xm
j=1
pj(t)x(j−1)(µj(t0k, t1k, t)), Λ(x)(t) = Xm
j=1
pj(t)x(j−1)(τj(t)). (2.30) Problems (1.1), (1.3) and (2.26), (2.27) are considered in the case
pj ∈Lloc(]a, b]) (j = 1, ..., m), q, qk ∈Le22n−2m−2,0(]a, b[), (2.31) and for an arbitrary (m−1)-times continuously differentiable function x:]a, b] →R, we set
Λk(x)(t) = Xm
j=1
pj(t)x(j−1)(µj(t0k, b, t)), Λ(x)(t) = Xm
j=1
pj(t)x(j−1)(τj(t)). (2.32) Remark 2.1. From the definition of the functions µj (j = 1, . . . , m), the estimate
|µj(t0k, t1k, t)−τj(t)| ≤
(0 for τj(t)∈]t0k, t1k[ max{b−t1k, t0k−a} for τj(t)6∈]t0k, t1k[ follows. Thus, if conditions (2.25) hold, then
k→+∞lim µj(t0k, t1k, t) =τj(t) (j = 1, . . . , m) uniformly in ]a, b[. (2.33)
Lemma 2.6. Let conditions (2.25)hold and the sequence of the(m−1)-times continuously differentiable functions xk :]t0k, t1k[→ R, and functions x(j−1) ∈ C([a, b]) (j = 1, . . . , m) be such that
k→+∞lim x(j−1)k (t) =x(j−1)(t) (j = 1, . . . , m) uniformly in ]a, b[ (]a, b]). (2.34) Then for any nonnegative function w∈C([a, b]) and t∗ ∈]a, b[,
k→+∞lim Zt t∗
w(s)Λk(xk)(s)ds= Zt t∗
w(s)Λ(x)(s)ds (2.35)
uniformly in ]a, b[, where Λk and Λ are defined by equalities (2.30).
Proof. We have to prove that for any δ∈]0, min{b−t∗, t∗−a}[,and ε >0, there exists a constant n0 ∈N such that
Zt t∗
w(s)(Λk(xk)(s)−Λ(x)(s))ds
≤ε for t∈[a+δ, b−δ], k > n0. (2.36)
Let, now w(t∗) = max
a≤t≤bw(t), and ε1 = ε 2w(t∗)
Pm j=1
Rb−δ
a+δ |pj(s)|ds−1
. Then from the inclusions x(j−1)k ∈ C([a+δ, b−δ]), x(j−1) ∈ C([a, b]) (j = 1, . . . , m), conditions (2.33) and (2.34), it follows the existence of such constant n0 ∈N that
|x(j−1)k (µj(t0k, t1k, s))−x(j−1)(µj(t0k, t1k, s))| ≤ε1,|x(j−1)(µj(t0k, t1k, s))−x(j−1)(τj(s))| ≤ε1
for t∈[a+δ, b−δ], k > n0, j = 1, . . . , m.Thus from the inequality
|Λk(xk)(s)−Λ(x)(s)| ≤ |Λk(xk)(s)−Λk(x)(s)|+|Λk(x)(s)−Λ(x)(s)| ≤2ε1
Xm j=1
|pj(t)|, we have (2.36).
The proof of the following lemma is analogous to that of Lemma 2.6.
Lemma 2.7. Let conditions (2.28)hold and the sequence of the(m−1)-times continuously differentiable functions xk :]t0k, b] → R, and functions x(j−1) ∈ C([a, b]) (j = 1, . . . , m) be such that lim
l→+∞x(j−1)k (t) = x(j−1)(t) (j = 1, . . . , m) uniformly in ]a, b]. Then for any nonnegative function w∈C([a, b]), andt∗ ∈]a, b], condition (2.35) holds uniformly in ]a, b], where Λk and Λ are defined by equalities (2.32).
Lemma 2.8. Let condition (2.25) hold, and for every natural k, problem (2.23), (2.24) have a solution uk ∈Celocn−1(]a, b[), and there exist a constant r0 >0 such that
t1k
Z
t0k
|u(m)k (s)|ds ≤r02 (k ∈N) (2.37)
holds, and if n = 2m+ 1, let there exist constants ρj ≥0, ρj ≥0, γ1j >0 such that ρj = supn
(b−t)2m−j
Zt t1
(s−a)pj(s)ds
:t0 ≤t < bo
<+∞,
ρj = supn
(b−t)m−γ1j−1/2 Zt t1
(s−a) pj(s)
λj(b, t0k, t1k, s)ds:t0 ≤t < bo
<+∞,
(2.38)
for t1 = a+b2 , (j = 1, ..., m). Moreover, let
k→+∞lim ||qk−q||Le22n−2m−2,2m−2 = 0, (2.39) and the homogeneous problem (1.10), (1.2) have only the trivial solution in the space Cen−1,m(]a, b[). Then nonhomogeneous problem (1.1), (1.2) has a unique solution u such that
||u(m)||L2 ≤r0, (2.40)
and
k→+∞lim u(j−1)k (t) = u(j−1)(t) (j = 1, . . . , n) uniformly in ]a, b[ (2.41) (that is, uniformly on [a+δ, b−δ] for an arbitrarily small δ >0).
Proof. Suppose t1, . . . , tn are the numbers such that a+b
2 =t1 <· · ·< tn < b, (2.42) and gi(t) are the polynomials of (n−1)-th degree, satisfying the conditions
gj(tj) = 1, gj(ti) = 0 (i6=j; i, j = 1, . . . , n). (2.43) Then for every natural k,for the solution uk of problem (2.23), (2.24) the representation
uk(t) = Xn
j=1
uk(tj)− 1 (n−1)!
tj
Z
t1
(tj−s)n−1(Λk(uk)(s) +qk(s))ds gj(t)+
+ 1
(n−1)!
Zt t1
(t−s)n−1(Λk(uk)(s) +qk(s))ds (2.44) is valid. For an arbitraryδ ∈]0, a+b2 [, we have
t1
Z
t
(s−t)n−j(qk(s)−q(s))ds
= (n−j)
t1
Z
t
(s−t)n−j−1Zt1
s
(qk(ξ)−q(ξ))dξ ds
≤
≤nZt1
t
(s−a)2m−2jds1/2Zt1
t
(s−a)2n−2m−2Zt1
s
(qk(ξ)−q(ξ))dξ2
ds1/2
≤
≤n
(t1−a)2m−2j+1−δ2m−2j+1
1/2||qk−q||Le22n−2m−2,2m−2 for a+δ ≤t≤t1,
Zt t1
(t−s)n−j(qk(s)−q(s))ds ≤n
(b−t1)2n−2m−2j+1−δ2n−2m−2j+1
1/2×
×||qk−q||Le22n−2m−2,2m−2 for t1 ≤t≤ b−δ (j = 1, . . . , n−1).
(2.45)
Hence, by condition (2.39), we find
k→+∞lim
t1
Z
t
(s−t)n−j(qk(s)−q(s))ds= 0 uniformly in ]a, b[ (j = 1, . . . , n−1). (2.46) Analogously one can show that if t0 ∈]a, b[, then
k→+∞lim Zt t0
(s−t0)(qk(s)−q(s))ds= 0 uniformly on I(t0), (2.47) whereI(t0) = [t0,(a+b)/2] fort0 <(a+b)/2 and I(t0) = [(a+b)/2, t0] fort0 >(a+b)/2.
In view of inequalities (2.37), the identities u(j−1)k (t) = 1
(m−j)!
Zt tik
(t−s)m−ju(m)k (s)ds (2.48) for i= 0,1; j = 1, . . . , m; k ∈N,yield
|u(j−1)k (t)| ≤rj[(t−a)(b−t)]m−j+1/2 (2.49) for t0k ≤t≤t1k (j = 1, . . . , m; k∈N), where
rj = r0
(m−j)!(2m−2j+ 1)−1/2 2 b−a
m−j+1/2
(j = 1, . . . , m). (2.50) By virtue of the Arzela-Ascoli lemma and conditions (2.37) and (2.49), the sequence {uk}+∞k=1 contains a subsequence {ukl}+∞l=1 such that {u(j−1)kl }+∞l=1 (j = 1, . . . , m) are uni- formly convergent in ]a, b[. Suppose
l→+∞lim ukl(t) = u(t). (2.51)
Then in view of (2.49), u(j−1) ∈C([a, b]) (j = 1, . . . , m), and
l→+∞lim u(j−1)kl (t) =u(j−1)(t) (j = 1, . . . , m) uniformly in ]a, b[. (2.52) If along with this we take into account conditions (2.25) and (2.46), from (2.44) by lemma 2.6 we find
u(t) = Xn
j=1
u(tj)− 1 (n−1)!
tj
Z
t1
(tj −s)n−1(Λ(u)(s) +q(s))ds gj(t)+
+ 1
(n−1)!
Zt t1
(t−s)n−1(Λ(u)(s) +q(s))ds for a < t < b,
(2.53)
|u(j−1)(t)| ≤rj[(t−a)(b−t)]m−j+1/2 for a < t < b (j = 1, . . . , m), (2.54) u∈Celocn−1(]a, b[), and
l→+∞lim u(j−1)kl (t) =u(j−1)(t) (j = 1, . . . , n−1) uniformly in ]a, b[. (2.55) On the other hand, for any t0 ∈]a, b[ and naturall, we have
(t−t0)u(n−1)kl (t) =u(n−2)kl (t)−u(n−2)kl (t0) + Zt t0
(s−t0)(Λk(ukl)(s) +qkl(s))ds. (2.56)
Hence, due to (2.25), (2.47), (2.55), and Lemma 2.6 we get
l→+∞lim u(n−1)kl (t) =u(n−1)(t) uniformly in ]a, b[. (2.57) Now it is clear that (2.55), (2.57), and (2.37) results in (2.40) and (2.41). Therefore, u ∈ Celocn−1, m(]a, b[). On the other hand, from (2.53) it is obvious that u is a solution of (1.1). In the case wheren= 2m,from (2.54) equalities (1.2) follow, that is,uis a solution of problem (1.1), (1.2).
Let us show that u is the solution of that problem in the case n = 2m + 1 as well.
In view of (2.54), it suffice to prove that u(m)(b) = 0. First we find an estimate for the sequence {uk}+∞k=1. For this, without loss of generality we assume that
t1 ≤t1k (k∈N). (2.58)
From (2.44), by (2.39) and (2.49), it follows the existence of a positive constant ρ0, independent of k, such that
|u(m+1)k (t)| ≤
≤ρ0+ 1 (m−1)!
Zt t1
(t−s)m−1Λk(uk)(s)ds +
Zt t1
(t−s)m−1qk(s)ds
(2.59)
for t1 ≤t≤t1k,and
||qk||Le22n−2m−2,2m−2 ≤ρ0, (2.60) for k ∈N.On the other hand, it is evident that
Zt t1
(t−s)m−1Λk(uk)(s)ds ≤
Xm j=1
Zt t1
(t−s)m−1pj(s)u(j−1)k (s)ds +
+ Xm
j=1
Zt t1
(t−s)m−1pj(s) µj(tZ0k,t1k,s)
s
u(j)k (ξ)dξ ds
(2.61)
for t1 ≤t≤t1k (k∈N).
Let, now m >1. From Lemma 2.2 and condition (2.37) we get the estimates Zt
t1
|u(j)k (s)|2
(b−s)2m−2jds ≤
t1k
Z
t0
|u(j)k (s)|2
(t1k−s)2m−2jds≤22mr02 (2.62) for t1 ≤t≤t1k (j = 1, . . . , m). Then by conditions (2.38) we find
Zt t1
(t−s)m−1pj(s)u(j−1)k (s)ds =
= Zt t1
1 (b−s)2m−j
∂
∂s
(t−s)m−1u(j−1)k (s) s−a
(b−s)2m−j Zs t1
(ξ−a)pj(ξ)dξ ds
≤
≤ 4mρj
b−a Zt
t1
|u(j−1)k (s)| (b−s)m−j+2ds+
Zt t1
|u(j)k (s)| (b−s)m−j+1ds
≤
≤ 4mρj
b−a hZt
t1
(u(j−1)k (s))2
(b−s)2m−2j+2ds1/2
+Zt
t1
(u(j)k (s))2
(b−s)2m−2jds1/2i
×
×Zt
t1
(b−s)−2ds1/2
≤ 2mmr0ρj
b−a (b−t)−1/2
(2.63)
for t1 ≤ t ≤ t1k (j = 1, . . . , m). On the other hand, by the Schwartz inequality, the definition of the functions µj and (2.4) it is clear that
µj(tZ0k,t1k,s) s
u(j)k (ξ)dξ
≤ 2m−j
(2m−2j−1)!!λj(b, t0k, t1k, s)Zt1k
t0k
|u(m)k (ξ)|2dξ1/2
≤
≤2mr0λj(b, t0k, t1k, s)
(2.64)
for t1 < s ≤ t1k (j = 1, . . . , m). Then by the integration by parts and (2.38), (2.64) we get
Zt t1
(t−s)m−1pj(s) µj(tZ0k,t1k,s)
s
u(j)k (ξ)dξ ds
≤
≤2mr0
Zt t1
∂
∂s
(t−s)m−1 s−a
Zs
t1
(ξ−a)|pj(ξ)|λj(b, t0k, t1k, ξ)dξ ds
≤2mr0×
×ρj Zt t1
∂
∂s
(t−s)m−1 s−a
(b−s)γ1j−m+1/2ds ≤2mr0ρj×
(2.65)